HoAFM: A High-order Attentive Factorization Machine for CTR Prediction

作者：

Highlights：

• The HoAFM model can learn high order feature interaction efficiently.

• The HoAFM model can encode high-order feature interactions into feature representations in an explicit and efficient manner.

• We highlight the varying importance of interactions via two bit-wise attention mechanism.

摘要

•The HoAFM model can learn high order feature interaction efficiently.•The HoAFM model can encode high-order feature interactions into feature representations in an explicit and efficient manner.•We highlight the varying importance of interactions via two bit-wise attention mechanism.

论文关键词：Factorization machines,High-order feature interactions,Attention mechanism,Deep neural network

论文评审过程：Received 15 March 2019, Revised 18 June 2019, Accepted 2 July 2019, Available online 22 July 2019, Version of Record 20 October 2020.

论文官网地址：https://doi.org/10.1016/j.ipm.2019.102076